Many processes control the observed beta diversity (among-patch or among-site variation) of communities. We focus on the effects of sampled population size per site. Differences in beta diversity due to differences in sample size can be viewed as either (1) a sampling artifact (different size samples from the same underlying population may lead to different estimates of underlying beta diversity) or (2) the outcome of ecological processes (bottom-up or top-down controls may lead to differences in beta diversity due to actual differences in the population size). The ability to control estimates of beta diversity for effects of sample size can be very helpful in distinguishing sampling artifacts, or effects of overall changes in local population density, from ecological processes such as restricted dispersal or environmental filtering.
We propose a simple rarefaction protocol, analogous to the procedure used in studies of alpha diversity, that produces estimates of beta diversity that are controlled for differences in sampled local population size among treatments. We use a series of simple simulations to explore the effects of sampled local population size, due either to sampling or ecological processes, on different beta diversity metrics. We also present the results of applying the beta-rarefaction protocol to several existing data sets on beta diversity.
Results/Conclusions
The effects of sampled population size differ considerably for density- (Manhattan) and incidence-based (Jaccard) beta diversity indices. For the Manhattan (density) index, increasing sample size always decreases measured diversity, as the sampling variance among populations decreases; the effects are strongest for small baseline sample sizes, when overall beta diversity is low, and when rare species contribute most to beta diversity. For the Jaccard (incidence) index, the patterns are considerably more complex. Measured beta diversity is maximal when the probability of species occurrence is around 50%. At very low densities, beta diversity is low because species are absent almost everywhere, while at very high densities, it is low because they are present almost everywhere. When communities are composed of species from a mixture of abundance classes, beta diversity generally decreases with sampled population size but can show one or more intermediate peaks as individual abundance classes reach their maximal measured beta diversity. Our case studies, taken from predator exclusion trials in marine systems, show both cases where changes in beta diversity persist after rarefaction and cases where changes in beta diversity can be ascribed entirely to changes in local population density.